Human of One Labs

Science should be ahead of the claims made in its name, not the other way around.

H[1] is a research intelligence engine. Ask a research question in plain words; get back how much confidence the published evidence supports, every judgment behind that rating, and the studies excluded and why.

H[1] is your company’s “science advisory board,” reached over MCP by the apps you already build.

The Human of One portal: a user asks “Does creatine improve cognition in healthy adults?” and the corpus answers Very Low certainty, 38% agreement, over 8 studies.

An LLM can explain science. H[1] lets it show its work.

Governed research intelligence for AI applications.

General-purpose AI models are remarkably good at finding patterns, summarizing information, and communicating in plain language. But they are not, by themselves, scientific evidence-evaluation and recommendation systems. They can combine studies that answer different questions, overlook important limitations, or present a confident answer without a visible path back to the evidence.

H[1] gives the AI a specialized evidence and intervention layer. It organizes research into a structured evidence graph connecting each question to its outcomes, populations, studies, risk-of-bias judgments, exclusions, and certainty level.

MCP is the connection between them. It allows Claude or another compatible AI application to ask H[1] for evidence while you continue interacting with the AI normally.

The LLM provides the conversation. H[1] provides the scientific structure, methods, and interventions.

Three layers. One trustworthy answer.

1

The LLM: the interface

Claude, ChatGPT, or your application understands the user’s question, communicates the answer, and decides when scientific evidence is needed.

2

MCP: the connection

MCP gives the AI a standard way to call H[1] during the conversation—without requiring the user to leave the application or learn a new interface.

3

H[1]: the evidence system

H[1] searches a structured evidence graph, evaluates the relevant body of research, and returns the certainty level, supporting studies, limitations, exclusions, and reasoning behind the result.

Every person is an N of one.

Science studies populations. Life happens one person at a time.

Peer-reviewed research estimates what tends to happen across a particular sample, under particular conditions. Its conclusions depend on who was studied, what was measured, the study design, sample size, statistical assumptions, and the uncertainty surrounding the result.

That evidence is essential. But it does not automatically tell us what will work for a particular person—with a particular biology, history, environment, goal, and moment.

The average is useful. It is not an individual.

H[1] begins by making the population evidence transparent and trustworthy: what the research supports, how certain it is, and where its limits lie. That is the evidence layer any genuinely personalized system must be built on.

A research intelligence engine that assesses published studies with gold-standard methods and reports the certainty of the evidence, one outcome at a time.

GRADE certainty is not a score out of anything. It is a level — High, Moderate, Low or Very Low — reached by starting from study design and moving for stated reasons. Every move is recorded with the feature of the evidence responsible for it.

Two Ways to Use Our Engine

MCP

Connect it to Claude or ChatGPT and ask in plain words.

Connect it →

SDK

Use our SDK to build the engine into your own app.

Ask about the SDK →

The kind of question it answers

Ask in the portal, then open Your corpus to read the full assessment — the GRADE derivation, every domain judgment and its reason, and the studies behind it.

Opening Your corpus in the portal and scrolling a full GRADE assessment: the derivation, the domain-by-domain reasons, the Summary of Findings, and the body of evidence with risk-of-bias.

Cold plunges and recovery

Does cold water immersion after training improve recovery?

Read the worked example →

Creatine and cognition

Does creatine supplementation improve cognitive performance?

Interval training and VO₂max

Does interval training improve VO₂max in healthy adults?

Passive body heating and sleep

Does hot-water bathing before bed improve sleep?

Small, and honest about it.

Every number here was read out of the running corpus, not transcribed. The method is finished and tested; the body of assessed research is early, and grows a question at a time.

1

question rated end to end, shown in full

10

assessed for risk of bias · 10 admitted to a body of evidence

No study has been excluded from a body yet, and when one is, that will be visible too: anything at High or Critical risk of bias is excluded by name and domain, and No information is deliberately not an exclusion — a bias nobody could rule out is not a bias ruled out, so the study stays in and the body is downgraded for it.

We re-run our own assessments and publish whether they agree. The body-level rating reproduced on every domain across three independent runs. The study-level judgments underneath it — the layer the body is computed over — reproduced on 88% of domains in our first measurement, with one disagreement serious enough that it would have changed which studies entered the body. How we measure it, and how to re-run it yourself →

Meet the Team

Giselle Requejo

Founder · research systems & product

Software engineer and published neuroscience researcher; formerly CTO of Syneurgy. Builds the assessment pipeline, the MCP surface and the standard's machinery.

Michael Mannino, PhD

Chief Science Officer

Neuroscientist, published researcher, and editor of the Springer textbook Performance Neuroscience. His work spans complex systems, human performance, cognition, and AI. At H[1], he leads the scientific architecture, translating rigorous evidence standards into transparent, auditable systems that both people and AI can use.

See what H[1] evaluates next.

Receive occasional updates when we assess a new research question, expand the evidence corpus, or make a meaningful change to the methodology. If you have asked about a particular topic, we will let you know when its assessment becomes available.